Detecting differences in vegetation among paired sites using pollen records

The 'Qualitative Assessment of Difference' method (QAD) is proposed to objectively detect differences in the relative abundance of vegetation between paired sites using pollen percentages. This method corrects for intertaxonomic differences in pollen productivity and neutralizes influences of background pollen on pollen representation of vegetation, using an inverse form of the Extended R-value model. We test the method using modern pollen-vegetation data from small hollows in northern Michigan (6 taxa; 45 sites) and from northwestern Wisconsin (7 taxa; 43 sites) in the USA. Compared with pollen percentages, the one-tailed Fisher Exact test shows that the QAD method significantly improves the accuracy of the results for all taxa. The rank order of sites based on QAD is significantly correlated to the rank order of sites based on a survey of vegetation composition surrounding the hollow for each taxon (Spearman's rank-order correlation coefficients; p < 0.001). We apply QAD to 9000-yr pollen records from two forest hollows c. 6 km apart in northern Michigan. Among seven taxa compared, Pinus, a taxon with well-dispersed pollen and a high pollen producer, often displays discrepancies in the direction of difference between QAD and the percentage-based method, demonstrating that pollen percentages alone do not always reflect differences in vegetation composition accurately. When several similarly sized sites are available in the same vegetation zone, QAD can objectively rank-order sites for individual taxa in relation to, for example, differences in soils, topography, patterns of species invasion, natural and anthropogenic disturbances, and other factors.